时间标签：数据日期 2026-11-05 · 生成时间 2026-11-05 05:30（Asia/Shanghai）｜ 检索窗口 PubMed edat 2026/11/02 – 2026/11/06 ｜ 情报发生日期 2026-11-05
研究问题
Genotype provides a baseline prediction of metabolic phenotype, but determining whether that prediction reflects a patient’s current metabolic state requires consideration of additional factors scattered across the EHR.；LLMs could extract these signals while preserving clinically relevant context, including whether an exposure is current or historical and when it occurred relative to changes in treatment or clinical response.The temporal dimension is particularly important because phenoconversion is dynamic (Faber and Fuhr, 2004).；As new information enters the record, the same approach could support reassessment of whether the patient’s current clinical context remains consistent with the phenotype predicted by genotype.This approach may also extend beyond searching for predefined causes of phenoconversion.
方法
Importantly, inflammation-mediated suppression of CYP expression has emerged as a major contributor to phenoconversion, particularly in patients with infectious diseases, autoimmune disorders, cancer, or critical illness (Scodellaro et al., 2026).；Consequently, metabolic capacity should be viewed as a dynamic phenotype rather than a fixed genetic trait.Mounting evidence suggests that phenoconversion is remarkably common in real-world clinical settings, particularly among patients exposed to polypharmacy (Patel et al., 2024).；Large cohort studies have demonstrated that consideration of concomitant medications and other non-genetic factors substantially increases the proportion of patients with clinically actionable pharmacogenomic phenotypes compared with genotype-based prediction alone (Patel et al., 2024).
核心结论
Collectively, these resources and consortia have established an increasingly standardized framework for translating pharmacogenomic evidence into prescribing recommendations.；Large cohort studies have demonstrated that consideration of concomitant medications and other non-genetic factors substantially increases the proportion of patients with clinically actionable pharmacogenomic phenotypes compared with genotype-based prediction alone (Patel et al., 2024).；Recent advances in artificial intelligence, particularly LLMs, have demonstrated considerable potential to extract clinically relevant information from heterogeneous datasets, summarize complex patient histories, identify drug-drug interactions, and support clinical decision-making through context-aware reasoning (Du et al., 2026).
局限
However, implementation remains uneven and is concentrated in particular drugs, specialties, institutions, and healthcare systems (Kabbani et al., 2023).；This limitation is consistent with the broader principles of precision dosing, in which genetic information represents one of several patient-specific determinants that may influence pharmacokinetics, pharmacodynamics, efficacy, and toxicity.；Recent evidence has identified phenoconversion as an important limitation of genotype-only predictions, and several authors have proposed integrating pharmacogenomic results with clinical and pharmacological information to support precision dosing (Mostafa et al., 2022; Nahid and Johnson, 2022; Scodellaro et al., 2026; Klomp et al., 2020; Shah and Smith, 2015; Cicali et al., 2021; Preskorn et al., 2013; Stingl and Viviani, 2025).
对本项目的价值
中等价值：属「免疫衰老与慢性炎症」方向；免疫 / 炎症与衰老栏目可用作本项目文献库与综述素材。 ｜ 价值分 2.6/5（★★★☆☆）
全文来源
PubMed PMC JATS（正文 38,460 字符，题名指纹 0.89，已过 PubMed PMCID 官方关联 + 题名指纹 + 正文长度归属校验）
原文题录
- 中文题名：大语言模型在解决药物基因组学表型转化中的潜在作用
- 英文标题：The potential role of large language models in addressing phenoconversion in pharmacogenomics.
- 期刊与标识：Frontiers in genetics · PMID 42979603 · DOI 10.3389/fgene.2026.1987469 · PMC PMC13730465